Notes based on a meeting with Angela Mclean on 15 October 2009.
Modelling4All's epidemic modelling components will be part of an exhibit at the Royal Society Summer Science Exhibition.
We plan to have 3 computers driven by 3 mice.
One will have the basic SRI model
Another will have one where schools are introduced (and various school closing policies)
Another will add age and/or behavioural change once someone is aware that they are infected
These will be playable games but also can be re-designed by users by adding new micro-behaviours. (Small set of these building blocks since we expect to only have a few minutes per user (less for those who just play the games)).
If there is reliable Internet then we could aggregate visitors results to visualise the variability of outcomes for the same model and initial conditions.
Could also print out a small card with the user's model URL to continue working on it.
More ideas on 20 November in preparation for the 23 November meeting:
We need to keep the interface VERY simple. Ideally drag and drop. Could have a few behaviours (with nice icons and internal graphics on the micro-behaviour page) that can be added to the population or the public health authority that correspond to collections of micro-behaviours. For example,
Choices of school closing policies:
1. When first confirmed case in the school, 2. When a percentage of the population being infected is reached, 3. When a percentage of the general population is infected.
Choices of school re-opening policies:
Only if herd immunity reached?
Different behaviours for kids:
1. Kids stay home, 2. Kids go out to play with friends (could have different percentages of each behaviour).
Different behaviours for parents:
1. Stay home with the kids 2. Go to work and leave kids with other families
Could produce statistics such as:
- Total number infected
- Peak percentage infected
- Peak change in number in infected
- Time of peak
- Estimated economic cost (of missed work as well as school closing)
More ideas in preparation for the 9 February meeting:
Could generate households based real demographic data.
Start with a too easy game of "Stop the Epidemic!" where there are only homes and schools and button for closing all schools. Epidemic stopped.
Can add another score that measures economic costs (to the default health measure(s)).
If work places are added then the game gets interesting.
Could have visitors add a few other buttons:
- Automatic closing of schools with the rules above (Switcher to choose the strategy)
- Stay home if ill (cost of advertising campaign)
- Add hospital?
Could optionally have 2-player version with 2 cities, 2 measures, and 2 sets of buttons. Could even have travel between them.
Could have drag and drop interface for adding behaviours. If not could have special widgets on page that make it very easy to add them to the right place.
Heavy use of icons. No sliders. Maybe no graphs?
Maybe should visualise the virus as something large that attaches to people.
9 February afternoon meeting with Howard, Arthur, and Fabio
Consider modelling fatalities — another aspect of the "score" — One problem with this is that Swine Flu fatalities are too few for this to matter much for small population simulations.
A suggestion box at the display to encourage people to think about what could be different.
A "leader board" that indicates previous results. Perhaps more practical is that we do experiments in advance and fold some of those statistics into the games. Perhaps the panel containing buttons for modifying the game should end with "hundreds more — suggestion them in the box". E.g. public transport versus walking/bicycling. Age-related differences. Viruses that persist in the environment for a while.
The game should be designed to take a limited amount of time — e.g. one minute which may be a week or two.
Could have a list of things to observe but constrain things to a maximum number.
In later email Fabio suggested using bars that go up and down to represent observations rather than graphs that may be hard for some visitors.
Should consider whether vaccines should be something that can be added to games.
More ideas from 8 March reported in the M4A blog.
Accumulated ideas recorded 17 March:
Base the cost of school closing on real data. London School of Hygiene did a macro-economic study of the economic costs
Base households on social data. Founds some UK census data but heard that Essex University Life Style Survey is good.
Possible "take away" messages:
- trade-offs (costs of choices)
- the role of randomness
- models are wrong but useful
- can experiment with epidemics on in silico
Give visitors a taste of the enormous design space
Should "plumbing" behaviours such as HIDE be made invisible?
URL on game applet page should be customised for EGM.
Geography options? E.g. layout of schools, workplaces, homes, etc.
Use link shapes to have nice (optional) display of transmission links
Applet itself could be customised. Play instead of Go, followed by Pause, by Start Over. No clock switch? Output area only if referenced?
Could limit the number of widgets that a user can add by keeping count and disabling buttons when limit exceeded.
See how 3D NetLogo view looks and consider using it?
Game should have a sentence or two at top explaining itself and what an infected person looks like, etc.
Think about how to support 'league tables'.
Notes from meeting with Angela on 19 March.
Need an explanatory screen. Probably two: one for the game composer page and another for the applet page where you can run the game. Both can use HTML to be concise with links to details.
Start with "nice" base parameter set. Made good progress today defining that.
No need for "infected slider" that defines the number of index cases.
Use dark grey for recovered (rather than the virus with a slash through it).
Speed things up by stop checking for encounters if recovered.
Get rid of clocked button (and turn off).
Start with one school and no adults.
Changes visitors can do include:
- add schools (2 variants: (i) students go to school closest to home (ii) random school assignment
- school closing policy (need to rethink the close when any student infected policy in the one school situation)
- infectious trails
- hand washing (only makes sense if infectious trails)
- voluntary quarantine (includes delay in becoming aware of one's infection)
- economic constraints
Link (and handout) that sends people to website
Have summer science exhibition specific page that welcomes them back and leads them on to explore all the features we decided to remove (e.g. adults, vaccinations, …)
Should link to SIR mathematical theory behind epidemics and how close it is to one school where students never leave.
Notes from meeting Ken and Angela on 25 March.
Start with simulation (no game elements). One school, no trails, no sliders, no adults, but graph is there. Parameters set to be a nice baseline. Encourage multiple runs. Graph can show multiple runs. Ideally use solid, circles, dots but NetLogo may force us to use different shades of the same base colour.
Can add trails and add hand washing (no point hand washing ad campaign if no trails).
Many schools. Who goes to which school (closest or some other policy). No family with different schools.
Closing schools. Manual at first. Policies as well (maybe). With symptom delay gets interesting.
"Spending money"
- shutting schools cost per school per day (and maybe also per family affected)
- buying adverts for hand washing per day
- buying adverts for self quarantine per day
When out of money can no longer intervene.
Adding pre-symptomatic infectiousness.
Adding adults and workplaces.
Could generate sequence numbers (since no privacy at an exhibit anyway) so easy to continue at home or challenge friends.
Need a URL for fetching old games by sequence number.
Advanced mode makes micro-behaviours into buttons, etc.
A guide (pdf?) for teachers of young children (say 10) and a very different guide for a-level students (and their teachers). Video tutorial as well.
Same day Ken, Howard, Arthur and Fabio met.
Could add a "Catch it, bin it" campaign option which reduces the virus trail left behind of those that follow the campaign.
Could use a different colour for the virus if it is on rather than in (so hand washing is visualised as accomplishing something by removing the outside virus before it has a chance to go "inside").
"Parameters" is too technical — "factors you can change" or "numbers you can change"
Should use a "proper legend" for the applet page showing what the different visualisations mean.
Should display time and day. Maybe even the date (starting 23 June ?)
Could have a switch for switching between "exhibition mode" and "expert mode" that hides lots of detail. r1763 accomplishes much of this.
Remove buttons for Download and Share in exhibition mode (and clocked switch).
Buttons should really be check boxes that can be toggled to do and undo them.